【作者】 楼俊钢； 江建慧；

【Author】 LOU Jun-Gang~+,JIANG Jian-Hui (Department of Computer Science and Technology,Tongji University,Shanghai 201804)

【机构】 同济大学计算机科学与技术系；

【摘要】 软件可靠性预测是典型的小样本预测问题.传统的软件可靠性建模方法实际上是通过解决一个难得多的问题作为中间步骤来解决一个相对简单的问题.考虑到人工神经网络、支持向量机以及相关向量机在可靠性预测中的成果经验,借鉴Vapnik提出的转导推理理论构建了可靠性预测的转导推理理论.在四个数据集上的实验结果表明,基于转导推理的软件可靠性模型在预测能力上较之一些广泛使用的NHPP类模型有明显的提高.更多还原

【Abstract】 The conventional two-stage way of software reliability prediction tries to solve a relatively simple problem by first solving a much more difficult one.Due to the common knowledge in software testing that early failure behavior of the testing process may have less impact on later failure process,we introduce transduction inference theory into software reliability prediction for the successful experience of ANNs,SVMs and RVM in software reliability prediction.The prediction accuracy of software reliability prediction models based on ANN,RVM,SVM and some NHPP model is compared.Experimental results show that models based on transduction theory could achieve a higher prediction accuracy.更多还原